Temperature Correction of Near-Infrared Spectra of Raw Milk
José Antonio Díaz Olivares,Stef Grauwels,Xinyue Fu,Ines Adriaens,Wouter Saeys,Ryad Bendoula,Jean-Michel Roger,Ben Aernouts
DOI: https://doi.org/10.26434/chemrxiv-2024-ls0j0
2024-09-18
Abstract:Accurate milk composition analysis is crucial for improving product quality, economic efficiency, and animal health in the dairy industry. Near-infrared (NIR) spectroscopy can quantify milk composition quickly and nondestructively. However, external factors, such as temperature fluctuations, can alter the molecular vibrations and hydrogen bonding in milk, altering the NIR spectra and leading to errors in predicting key constituents such as fat, protein, and lactose. This study compares the effectiveness of Piecewise Direct Standardization (PDS), Continuous PDS (CPDS), External Parameter Orthogonalization (EPO), and Dynamic Orthogonal Projection (DOP in correcting the impact of temperature-induced variations on predictions in milk long-wave NIR spectra (LW-NIR, 1000 to 1700 nm).
A total of 270 raw milk samples were analyzed, collecting both reflectance and transmittance spectra at five different temperatures (20°C, 25°C, 30°C, 35°C, and 40°C). The experimental setup ensured precise temperature control and accurate spectral measurements. PLSR models were calibrated at 30°C to predict milk fat, protein, and lactose content. The performance of these models was assessed before and after applying the temperature correction methods, with a primary focus on reflectance spectra.
Results indicate that EPO and DOP significantly enhance model robustness and prediction accuracy across all temperatures, outperforming PDS and CPDS, especially for lactose prediction. These orthogonalization methods were compared against PLSR models calibrated with spectra from all temperatures. EPO and DOP showed comparable or superior performance, highlighting their effectiveness without requiring extensive temperature-specific calibration data. These findings suggest that orthogonalization methods are particularly suitable for in-line milk quality measurements under farm conditions where temperature control is challenging. This study highlights the potential of advanced chemometric techniques to improve real- time, on-farm milk composition analysis, facilitating better farm management and enhanced dairy product quality.
Chemistry